本文へ移動
cccskills
無料GitHub で公開

token-broker-client

Brokered short-lived provider credentials client in matrx-frontend (`lib/api/broker/`). Use when browser code needs a provider realtime session, a direct provider call, an ephemeral token, or a provider API key in the browser or `NEXT_PUBLIC_*`. NOT for adding a new audience server-side (use aidream's token-broker).

インストール方法を見る

含まれるファイル(1)

  • SKILL.md3.1 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

token-broker-client — consume brokered credentials

The primitive lives at lib/api/broker/ (contract: its FEATURE.md; cross-repo system of record: ../common-docs/systems/architecture/token-broker/FEATURE.md). It is complete — consume it; never hand-roll a mint call, credential cache, or gateway fetch.

Decision gate — does the server support the audience?

Check KNOWN_AUDIENCES in lib/api/broker/types.ts and the live minter registry (aidream services/token_broker/minters/).

  • Supported → consume through the primitive (below). Zero broker-infrastructure code.
  • NOT supported (new provider/capability) → STOP client-side. The system grows server-first: follow the global token-broker skill (~/.claude/skills/token-broker/SKILL.md) — one minter + one registry line in aidream — then consume here through the unchanged primitive. Never work around a missing audience with client-side key handling.

How to consume

import { brokeredFetch, resolveBrokeredTransport } from "@/lib/api/broker/transport";
import { useBrokeredCredential } from "@/lib/api/broker/hooks";

// Proxied audience (e.g. "anthropic") — provider-wire HTTP, streaming included:
const res = await brokeredFetch(
  { audience: "anthropic", tierPolicy: "none", model: "claude-sonnet-5" },
  "/v1/messages",
  { method: "POST", body: JSON.stringify(anthropicShapedBody) },
);

// Native audience (e.g. "openai_realtime") — connect directly per protocol:
const t = await resolveBrokeredTransport({ audience: "openai_realtime", tierPolicy: "none" });
if (t.mode === "native_ephemeral") connectRealtime(t.endpoint, t.token);

// Component that needs a continuously-fresh credential:
const { credential, status, error, refresh } = useBrokeredCredential(req);

Caching, in-flight dedup, refresh-ahead (<20% TTL), 401→one-re-mint, and sign-out clearing all happen inside the primitive — callers never manage credential lifetime.

Non-negotiables

  • tierPolicy is explicit at every callsite ("none" | "guest" | "mid"). Never add a default, wrapper-with-default, or omittable param.
  • Never persist or log a token. Memory only. Never render credential.token — mask it (debug UIs use useBrokerCacheSnapshot, masked).
  • Never hardcode a gateway/provider URL — credential.endpoint is the only source.
  • No provider key in client env (NEXT_PUBLIC_* or otherwise). Finding one is a defect: replace it with a broker audience (server-first).
  • 503 on mint = loud error ("broker unconfigured"), never a silent fallback auth path, never a retry loop.

Verify

Demo page /demos/token-broker (dev builds): mint per audience, cache inspector, live proxied Anthropic call, forced re-mint.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Admin debug system wiring for the floating AdminIndicator's live debug data and Copy Full Context. Use when adding debug visibility to a route or feature, wiring useDebugContext or the debug panel, capturing console errors, or enabling the copy-context workflow for a page.

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

Compact two-icon Copy / Copy-for-AI controls (components/agent-copy). Use when adding copy buttons to a row, card, list, or record; merging duplicate Copy/JSON/AI controls; continuing the copy rollout; or writing a Copy-for-AI payload. NOT for markdown content actions (use rich-document-actions).

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

Disclosing a surface's existing fixed AI jobs in the shell's top Agents menu. Use when a page, panel, overlay, or window already runs a mandate behind a button, assist, automatic action, or mode; when the agent-disclosure guard names a file; or during a surface check.

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

The watch-fix-rerun method for making an existing platform agent or automated process efficient and correct: baseline its ledger, run one unit yourself, fix the class behind every wasted call, rerun, record. Use when asked to improve, optimize, watch, or 'make efficient' an agent, a sandbox session, a sync job, or any recurring automated process.

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

Redux state for agent execution and firing agent shortcuts. Use when editing features/agents/redux/, building agent UI, creating a conversation, touching assembleRequest or NDJSON stream state, wiring agent-state selectors, adding a per-conversation capability, or triggering a shortcut from a button, menu, or code (useShortcutTrigger, launchAgentExecution).

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

The provision for a call site: the exhaustive menu of values that place in the code can realistically produce. Use before creating or fixing a mandate or agent that reasons about more than its own input, when deciding whether an agent can answer at all with what it is sent, or when its output 'looks right' but nobody checked it could be. NOT for building the agent (use create-agent).

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

armanisadeghi のスキルをすべて見る

このスキルの問題を報告する